The CABG Patch Trial is testing the hypothesis that prophylactic use of implantable cardiac defibrillators (ICDs) will improve survival in high-risk coronary heart disease patients undergoing CABG surgery. The original design called for 800 patients to be randomized to ICD prophylaxis or to no therapy and followed for 2 to 6.5 years (average, 40 months) to a common termination date. Since the ICD pulse generators used in this trial lasted about 42 months, the original design required ICD replacement in many patients. At its first two meetings in 1993, the Data and Safety Monitoring Board (DSMB) formalized a plan to adjust sample size in October 1994 if the control group mortality rate was lower than expected. In June 1994, an unanticipated and unique event—a subpoena from the Office of the Inspector General (OIG)—made it impossible to replace about half of the ICD generators and threatened to shorten follow-up substantially. If follow-up had been stopped on the date originally planned, but without replacing ICDs, the average follow-up would have fallen from 40 months to about 33 months. Also, in October 1994, the control group mortality rate was found to be somewhat lower than expected. Together, the abbreviated follow-up and lower control group mortality threatened to reduce power substantially. The DSMB reviewed several options for restoring power. Because mortality rates in the first month after CABG surgery were about seven times as high as thereafter and because ICD therapy did not reduce surgical mortality (death during the first 30 days), extending the follow-up benefits power more than does increasing the sample size. However, the limit on extending follow-up was 42 months (the expected battery life of the ICD). Data from the ICD-treated group was not reviewed or considered in making the decision. After reviewing many options for restoring power, the DSMB recommended that the sample size be increased from 800 to 900 patients and that almost all patients be followed for 42 months. This recommendation extended follow-up for 2 years beyond the original termination date planned for the trial and dictated that patients close out after 42 months rather than on a common termination date.